Zeroing SCACP VC/CHANNEL error counts?

[Graff]

I had some problems on our internal network caused by misconfiguration of some of the interfaces in our ten-node V8.3 Alpha cluster.  These misconfigurations are, I believe, now fixed.  I would like to confirm this belief by checking the error counts in the SHOW VC/ALL and SHOW CHANNEL/ALL output from SCACP.  Unfortunately, the error counts displayed by these commands are large.  I would like to zero those error counts, so that in 24 hours when I rexamine the counts, I see "20" and don't have to compare two numbers in the millions (!), but I don't know how (to zero the counts, not subtract two numbers...).  I know about the SET DEVICE/RESET=ERROR_COUNT command, but that didn't reset the SCACP error counts.  I scanned the "New Features and Documentation Overview" volumes for the past few OS releases, but didn't see anything relevant.  Rebooting the cluster isn't an option.  Is there a documented way to reset those counts?

Gareth

[John Gillings]

Gareth,

   I don't believe there's a documented way to reset the counters. There's probably a way you could do it by hacking system data structures (not recommended). You wouldn't have believed the trouble it took to get SET DEVICE/RESET=ERROR_COUNT implemented! There are still folk who consider it an abomination.

  You could write some code to parse the output, but another possibility is to use the /INTERVAL qualifier to display deltas as well as absolute counts. If you don't want to leave an interactive SCACP session open for a long time, try something like the attached procedure. The idea is to generate a sequence of commands at regular intervals, piped into a process running SCACP. Run it as a batch job, the log file will contain timestamped samples taken every 10 minutes, with the absolute number and the delta. Use parameters to adjust interval and/or number of samples. Small matter of text processing to convert it into a T4 file.

[Graff · Accepted solution]

Thanks for the suggestion.  I'm running the command file now.

When I first tried the /INTERVAL flag, I didn't get any deltas.  I have since realized that the flag doesn't work over separate invocations of SCACP.  I used SYSMAN to run SHOW VC/COUNT across the cluster, capturing the output to a file.  Then, sometime later, I reran the command.  Separate invocation of the SCACP command on each machine, so no deltas :-(  The on-line documentation for /INTERVAL might benefit from a little further explanation.

I think my best bet is to write a short command file to invoke SYSMAN once a night, run the SCACP SHOW VC and SHOW CHANNEL commands, write the output to a file, and maintain "tonight's" file and "last night's" file.  A simple program can then produces the delta values from the two files.

Gareth